ORGANIZATIONAL STRUCTURE AND PERFORMANCE FEEDBACK:
SITUATED DECISION MAKING AND PERSISTENCE IN PRODUCT PHASE-OUT
John Joseph
Duke University
100 Fuqua Drive
Durham, NC 27718
p. 919-660-4083
e. john.e.joseph@duke.edu
Ronald Klingebiel
Warwick Business School
Coventry, CV4 7AL
United Kingdom
p. +44 (0)24 7652 4622
e. ronald.klingebiel@wbs.ac.uk
Alex James Wilson
Duke University
100 Fuqua Drive
Durham, NC 27718
p. 919-381-2055
e. alex.j.wilson@duke.edu
February 1, 2013
We thank Rich Bettis, Rich Burton, Vibha Gaba, William Ocasio and participants at the Academy of Management
symposium, Wharton Seminar Series and Duke Seminar Series for their helpful comments on previous versions of
the paper.
ORGANIZATIONAL STRUCTURE AND PERFORMANCE FEEDBACK:
SITUATED DECISION MAKING AND PERSISTENCE IN PRODUCT PHASE-OUT
Abstract
This study examines the effects of organizational structure on product phase-out. Using quarterly
product-level data on the top five mobile handset manufacturers for the period 2004–2009, we
analyze how the elevation of phase-out decision to higher levels in the firm—and how the extent
of consultation at that level—influence persistence in phase-out decisions. Results show that
elevation speeds phase-out whereas consultation slows phase-out. However, structure also
moderates the effect of persistence following positive and negative feedback. These results
suggest that elevation can increase the inertia and bias in phase-out decisions but only when the
firm is well above (or well below) aspirations. Our findings suggest also that the bargaining
prevalent in consultative environments facilitates phase-out when performance is below
aspirations, but may induce persistence when above. A broader contribution is to offer a theory
of situated selection according to which organizational structure affects phase-out through
information processing and the way managers attend and respond to performance feedback.
Keywords: organizational structure, performance feedback, persistence, product exit
1
In high-technology industries, rapid technological change requires that firms actively
manage their product portfolio by carefully timing the introduction of new products and the
selective removal of old ones (Burgelman, 1984, 1994; Henderson and Stern, 2004). Studies
confirm that new product introductions help organizations diversify and reinvent themselves
(Schoonhoven, Eisenhardt, and Lyman, 1990) and may also improve chances of survival
(Banbury and Mitchell, 1995). No less important, though far less studied, is product phase-out or
culling: the internal decision to withdraw a product from the market. The phase-out of older
products and resulting turnover of product portfolios helps the producer adapt to changing
market and technological conditions, which is a critical factor in the firm’s success and evolution
(Sorenson, 2000; Burgelman, 2002; de Figueiredo and Kyle, 2006). The timing of product phaseout affects firm revenues as well as the allocation of resources and managerial attention to
products in the portfolio (Henderson and Stern, 2004).
Phase-out decisions may be especially challenging when the firm has been exceptionally
successful or unsuccessful. Theories of behavioral persistence demonstrate that organizations
tend to persist with actions previously associated with desirable outcomes (Lant, Milliken, and
Batra., 1992; Miller and Chen, 1994; Audia, Locke, and Smith, 2000; Guler, 2007), a behavior
that tilts the balance toward leaving products on the market too long. Organizations are likely to
repeat actions associated with favorable feedback (Cyert and March, 1963; Prahalad and Bettis,
1993; Greve, 1998; Burgelman, 2002) because success creates overconfidence in existing
knowledge and thereby limits the search for new alternatives (Levinthal and March, 1993).
Owing to this “paradox of success,” organizations often remain committed to old products and
neither respond to technological or environmental change nor develop new capabilities (Audia,
Locke, and Smith, 2000). Similar behavior may be observed following poor performance.
2
Although it seems counterintuitive to retain poorly performing products, several studies
demonstrate that legacies may persist when organizations struggle to interpret properly—and act
on—the feedback from negative performance (Cannon and Edmondson, 2001; Baumard and
Starbuck, 2005; Rerup, 2009). Learning from negative feedback may be difficult (Audia and
Greve, 2006; Eggers, 2012), which helps explain why a firm may escalate its commitment (Staw,
Sandelands, and Dutton, 1981; Guler, 2007) or discount alternatives that, given limited initial
information, appear not to be viable (Denrell and March, 2001). For example, Burgelman (1994)
found that, despite lagging performance in their memory chip business, Intel executives had
difficulty switching their focus from memory to microprocessors; similarly, Guler (2007)
demonstrated that venture capital firms are often reluctant to terminate even clearly unsuccessful
investments.
Although there is robust support for models of behavioral persistence, this body of work
generally assumes that decisions—such as those concerning product life—are handled uniformly
throughout the firm. That assumption fails to acknowledge managers as the primary determinants
of whether individual strategies and products are eliminated or selectively retained (Burgelman,
1991; Henderson and Stern, 2004), and it reflects the view that performance feedback is
invariantly processed regardless of the decision maker’s position within the corporate hierarchy.
However, studies grounded in the behavioral theory of the firm indicate that responses to
feedback may vary with the subunit in which evaluations of success and failure occur. For
example, Gaba and Joseph (2012) examined the unique effects of corporate and business unit
performance feedback on new product introductions and found different effects for each. Audia
and Sorenson (2001) similarly concluded that different functional areas create and monitor their
own unique performance metrics and react differently to their respective feedback. These studies
3
focus on responses to different feedback within the firm, so they leave open the question of how
the same feedback is processed by different organizational structures. The organization’s
decision-making structure is therefore a significant omission in studies of behavioral persistence
and of performance feedback more generally. We define organizational structure in terms of
explicitly mandated formal structure and also in terms of formal and informal interactions among
individuals in different functions or units (cf. Gulati, Puranam, and Tushman, 2012).
How does an organization’s decision-making structure affect persistence behavior? To
answer this question, we set our study in the mobile device industry and examine, across the
largest firms, the problem of product phase-out. Organizational structure may have important
implications for persistence in product phase-out within large vertical hierarchies because it
shapes not only the efficient processing of information, which affects coordination of activities,
but also attention to and perceptions of information (e.g. feedback), which has implications for
how firms respond. Structure’s role in information processing is a well-established one, in that
the firm’s decision making structure enables the efficient collection, processing and distribution
of information to help managers navigate and address demand uncertainty (Tushman and Nadler,
1978). Less observed is how organizational structure affects the attention and responsiveness to
performance feedback. Studies have shown that decision making is situated (Lave and Wenger,
1991; Ocasio, 1997; Elsbach, Barr, and Hargadon, 2005), and the organizational structure itself
has been shown to affect decision making and performance (Csaszar, 2012; Csaszar and Eggers,
2013) as well as cognition and the development of capabilities (Tripsas and Gavetti, 2000;
Gavetti, 2005). Accordingly, the structural location of decision makers may amplify or attenuate
the inertial properties of and biases in decision making after success or failure. Little is known
about this aspect of structure, so explorations along these lines answer the call of scholars to
4
reintegrate structure into the behavioral foundations of the Carnegie tradition (Gavetti, Levinthal,
and Ocasio, 2007).
In light of this important lacuna, we examine the effects of two key structural factors: the
elevation of decisions up the hierarchy, and the extent to which portfolio decisions involve
consulation with peers. Our main thesis is that structure affects phase-out through information
processing in support of coordination and that, because structure situates each decision maker
within a particular subenvironment, it also shapes how managers respond to feedback. We argue
that a manager’s response to success and failure—specifically, to performance above and below
aspirations—may differ because phase-out decisions involve weighing the effects of lost sales
(which result from pulling a product too early) against the effects of “cannibalization” (the
process by which a new product gains sales by diverting them from the firm’s existing products).
The manager’s appraisal will vary with the change in attention focus that accompanies elevated
decision making and the negotiations required under a consultative decision-making regime.
We test our predictions using a unique data set of product sales in the German mobile
device industry. This industry is a suitable setting because it is characterized by a high level of
product turnover (i.e., older products are quickly replaced by newer ones). Henderson and Stern
(2004) observed that, “although eliminating a viable, revenue-producing product seems
counterintuitive, it is quite common in high-velocity settings, in which obsolescence occurs
quickly” (45). Timely phase-out ensures that older products will not divert managerial attention
and resources from new products (Greenstein and Wade, 1998). For handset makers, key
performance indicators (e.g., unit sales) are widely shared and are watched closely by industry
analysts and investors. The single-industry and single-country context of our study minimizes the
risk of unobserved heterogeneity because all handset makers are engaged in similar production
5
activities. The study’s observations are from a period during which overall portfolio health,
rather than “hit phone” status, were firm’s primary performance benchmark.
In addition, the organizational charts of our sample firms are roughly similar. Mobile
device firms typically give product managers responsibility for a series of products. For example,
at Motorola the mobile devices business was subdivided into units dedicated to different
technological standards (e.g. CDMA, GSM, iDEN) and into teams managing various smartphone
platforms. In each major firm there is a business unit head as well as several layers between that
head and the product managers. The firms in our sample also featured such common functions as
research and design, product marketing, and sales. However, the process of making decisions
about the product portfolio vary; some firms vest such decisions at the product manager level
whereas other firms elevate the decision responsibility upward in the organizational hierarchy.
Our study makes several contributions to the literature. First, we contribute to theories of
behavioral persistence by positing a role for organizational structure in product phase-out
decisions. For this, we link theories of situated decision making (Ocasio, 1997; Elsbach, Barr,
and Hargadon, 2005) and performance feedback (Lant, Milliken, and Batra, 1992; Audia and
Sorenson, 2001; Greve, 2003b; Gaba and Joseph, 2012) to develop a theory of situated selection
in product phase-out. Our theory suggests that internal selection—that is, selective removal of
products from the market—is guided by the locus of decision making (elevation) and local
interactions (consultation), which underscores the heterogeneity of selection processes within
and between firms. Second, we augment theories of performance feedback by elaborating the
role of organizational structure. We demonstrate that actions carried out in response to feedback
are also function of the organizational level of decision making and the extent of consultation at
that level. More generally, our findings of situated selection offer new insights into how
6
structural and cognitive drivers interact to shape adaptation, of which product phase-out is but
one example.
ORGANIZATIONAL STRUCTURE AND DECISION MAKING
For much of the past five decades, research on organizational structure has centered on two
critical design choices (Simon, 1962). The first of these choices concerns centralization versus
decentralization of decision making within the firm (Egelhoff, 1982; Miller and Droge, 1986). In
other words: Should decisions be made at lower or higher levels of the organizational hierarchy?
The second design choice concerns the extent to which individual units should function
autonomously or rather collaborate in decision making (Lawrence and Lorsch, 1967). These two
dimensions are especially salient for large firms, which are generally characterized by a system
of subunits and multi-level hierarchies that collectively allocate resources while formulating
business policies and strategic plans (Chandler, 1962; Williamson, 1985).
Research that addresses the decision-making implications of organizational structure has
traditionally been anchored by information processing theory: the study of how structural choices
bear directly on the firm’s capacity to collect, process, and distribute such information as plans,
budgets, market conditions, and performance feedback (Tushman and Nadler, 1978: 614; see
also Galbraith, 1974). From this perspective, the role of structure is to increase informationprocessing capacity—both vertically and horizontally—and thereby improve coordination among
various organizational functions (Gulati, Lawrence, and Puranam, 2005).
We speak of vertical information processing when referring to the elevation of portfolio
decisions within the vertical hierarchy. In some firms, formal roles and responsibilities or
standard operating procedures may lead to these decisions being pushed up the chain of
command through such conduits as cross-level communication channels (Galbraith, 1974;
7
Joseph and Ocasio, 2012). Elevation may also follow routines and unwritten rules of engagement
transmitted through social networks and interactions that are less formal. The logic of this
argument is that, whether formalized or not, more efficient information processing leads to better
outcomes—especially when the environment is complex and/or changing rapidly (Siggelkow and
Rivkin, 2005).
By the term horizontal information processing we reference the degree to which
information flows between actors at a given level of the organizational hierarchy. Recall that
Lawrence and Lorsch (1967) established both the differentiation and the integration of a firm’s
information-processing capacity as critical determinants of performance. We focus here on the
integration that occurs through consultation, defined as communication and negotiation between
actors—via formal and/or informal interactions—that aims to achieve consensus concerning
portfolio decisions. Consultation may occur through various information-processing linkages,
including dedicated roles, cross-functional teams, organizational goals, and informal channels
(Galbraith, 1974). Such consultation is consequential for both cooperation and coordination with
regard to a wide variety of activities (Gulati, Lawrence, and Puranam, 2005).
However, information processing is only one mechanism through which organization
structure (here, elevation and consultation) may affect decisions such as product phase-out.
Organizational structure may also have an effect through its moderating effect on the attention
to, interpretation of, and response to information—in particular, performance feedback.
Feedback-based response is the attempt by managers to understand the connections between their
actions and the organization’s outcomes, and it involves responding to performance that is
compared to a reference point or aspiration level. From this perspective, the role of structure is
not to facilitate information-processing capacity so much as to determine whether and how the
8
firm attends and responds to feedback. A given firm’s organizational structure creates
idiosyncratic decision-making contexts within subunits, which leads attention to be focused on
certain aspects of feedback information to the exclusion of other aspects (Ocasio, 1997).
In short: responses to performance feedback are not uniform, and the cognitive processes
underlying any phase-out decision vary according to whether or not decisions are elevated and
whether or not they involve consultation with multiple decision makers. Therefore, if decisionmaking structure varies across organizations then responses to comparable feedback—and
choices concerning phase-out—may well differ from one decision maker to the next. In what
follows, we examine both information-processing and feedback-related effects of structure.
Elevation of phase-out decisions. The elevation of portfolio-related decisions has several
implications for the speed of phase-out. With elevation, decisions are transferred vertically from
a product manager to a more senior-level decision maker—for example, a vice-president or the
business unit chief. Because elevation naturally vests decisions with a more centralized manager,
higher-quality and more diverse knowledge can be brought to bear on phase-out decisions.
Centralization helps to facilitate dense internal communication flows and increase firm
absorptive capacity (Jansen, Van Den Bosch, and Volberda, 2005). Vertical communication is
more efficient than horizontal communication because subunit managers are more willing to
share information with centralized decision makers. Such information is less likely to be biased
in that competing product teams prefer vertical to horizontal communication (so they can
monopolize insights that may advantage them (Alonso, Dessein, and Matouschek, 2008). Better
information quality will likely yield faster decisions as senior managers are then no longer
required to seek out qualifying information or second opinions or to sort through mixed
messages.
9
Centralization also limits the ability of product managers to assert parochial agendas. Product
managers tend to screen out information that does not affect their particular products,
technologies, or markets; they naturally focus on the problems, actions, and outcomes that reflect
their role within the firm (Henderson and Cockburn, 1994). A product manager is probably not
aware of all the resource demands of other product managers and does not know (or care) exactly
where resources should be directed, an ignorance that hampers coordination efforts meant to
support phase-out. Higher-level managers are not embedded in local information filters Argote
and Ingram, 2000); hence they can access and appreciate information concerning the entire
portfolio and are therefore better positioned to facilitate phase-out. Because these executives
consolidate and adjudicate agendas held by lower-level managers, they are better positioned also
to manage the entire range of demands across subunits and thus, once again, can more readily
and accurately identify phase-out candidates.
By extension, centralized decision makers may be better able to facilitate knowledge
exchange across subunits and hence to prevent phase-out decisions that could have mutually
destructive consequences (Siggelkow and Rivkin, 2005). Because phase-out involves adjusting a
range of interdependent activities—such as factory schedules, operator road maps, and supply
chains—greater access to the complete spectrum of information improves coordination in
support of product discontinuation. Overall, the efficiency of information processing improves as
decisions are elevated within the organization, which suggests the following hypothesis.
Hypothesis 1 (H1): The likelihood of phase-out increases with the elevation of product
portfolio decisions.
Consultation in phase-out decisions. Consultation has implications both for unity of effort
(cooperation) and for unity of action (coordination) in support of phase-out (Lawrence and
Lorsch, 1967; Gulati, Lawrence, and Puranam, 2005). Problems of cooperation arise from
10
conflicts of interest. Under assumptions of opportunism or self-interest, collective agreement
may fail to occur because of actions motivated by the private benefits of managers. To deal with
these inefficiencies and to motivate managers in a uniform way, organizations often put in place
incentives, sanctions, and monitoring mechanisms (Williamson, 1985). They may also rely on
more informal mechanisms—in particular, consultation (interactions) between units and
managers. Frequent consultation may improve cooperation because it facilitates the creation of
trusting relationships and social ties (Gulati 1995; Tsai and Ghoshal, 1998). Interactions create
an environment in which a manager can develop a common shared purpose, and they also
establish arenas for sharing information and engaging in collective action (Martin and
Eisenhardt, 2010; Martin, 2011). Consultation thus facilitates the alignment of interests among
managers responsible for different product groups or functions and, at the same time, promotes
collectively agreed-upon product life cycles and phase-out dates.
Yet despite the cooperative advantages that result from greater information sharing
(Huber, 1991), consulting with colleagues in the context of portfolio management may, on
balance, actually slow down phase-out owing to the additional coordination costs that arise from
multiple interests. The need for consensus requires that product phase-out decisions be approved
by others, which means that diverse and often conflicting demands must be served by any given
phase-out decision (Cyert and March, 1963). This negotiation process takes time—and more than
that required for the mechanism of H1, whereby decisions are referred up the organizational
hierarchy but not necessarily discussed at length. Furthermore, efforts to establish consensus may
lead to open disagreement about the means necessary to accomplish objectives or, more
fundamentally, may create inconsistency in aims; these effects may complicate or delay decision
making (Denis et al., 2011)
11
For example, product portfolio decisions at Motorola in the mid-2000s involved
managers from different functions. Thus portfolio meetings would include representatives from
R&D, design, marketing, and production as well as managers handling the relationships with
Vodafone, T-Mobile, and other major telecom operators. Decisions to launch or cull products
were committee decisions that required a relatively high degree of consensus. On the one hand,
this ensured that product decisions were informed by better information; on the other hand, it
continually forced managers to revisit and compromise on timing that was appealing to no one
but at least acceptable to everyone. The approach was described by one manager as a
“redecision-making democracy” that delayed portfolio management decisions of all kinds. So
consultation has both advantages and disadvantages. Although it achieves the pooling of
necessary information, it may also end up inhibiting action. On balance, we suggest the
following hypothesis.
Hypothesis 2 (H2): The likelihood of phase-out decreases as the extent of within-level
consultation increases.
Organizational structure and portfolio feedback. Organizational structure may affect phaseout not only through information processing but also through its effect on how the firm responds
to performance feedback information. This is because managerial responses to success and
failure may differ with variations in the attention-directing qualities of the organizational
structure (Ireland et al., 1987; Milliken and Lant, 1992; Ocasio, 1995). In particular, we argue
that the effect of performance feedback on product phase-out varies with both the elevation (up
the hierarchy) of portfolio management decisions and the extent of within-level consultation.
Our feedback model is grounded in the behavioral theory of the firm. In this model,
boundedly rational decision makers simplify performance evaluations by transforming a
continuous measure of performance into a discrete measure of success or failure (March and
12
Simon, 1958; Cyert and March, 1963; March, 1988). To do so, decision makers evaluate
performance—on some key dimension—with respect to an aspiration level, reference point, or
goal (Miller and Chen, 1994; Greve, 1998; Audia, Locke, and Smith, 2000; Mezias, Chen, and
Murphy, 2002). The difference between performance and aspirations serves as a feedback
mechanism, which provides a signal to decision makers regarding whether or not to maintain
current activities and/or to adjust aspiration levels (Greve, 2003b). The aspiration level thus
serves as the dividing line between perceived success and failure, and a firm’s most recent
performance may be the starting point for a revision of phase out decisions. Although the effect
of portfolio feedback on product exit has not been a topic of prior studies, the theory suggests
that past portfolio performance may affect product persistence owing to biased managerial
perceptions concerning the success or failure of the unit’s portfolio performance (Milliken and
Lant, 1991).
Successful organizations tend to experience more inertial pressure than do less successful
ones because the pressure to remain consistent is naturally much stronger after a period of
success than of failure. External stakeholders will likely expect a continuation of activities that
have yielded success in the past, so managers are prone to favoring the status quo and thus to
extending product life. Success serves as an indicator that the firm’s actions are effective and
often leads to the selective retention of activities that contributed to that success (Greve, 1998;
Kraatz, 1998; Audia, Locke, and Smith, 2000; Schwab, 2007). Successful performance can make
managers “so complacent, so content with the status quo, that they resist change” (Miller and
Chen, 1994: 3). Signals of past success may encourage managers to retain existing products
longer, even when faced with technological or environmental change or the availability of newer
products to replace them (Lant, Milliken, and Batra, 1992).
13
Forces inducing persistence may also play a role in responses to failure. When
performance does not meet aspirations, managers may escalate their commitment to failing
products—making decisions that disregard negative feedback about prior resource allocations,
that discount the uncertainty of goal attainment, and that result in the inability to break from a
current course of action because of the psychological sunk costs associated with doing so
(Brockner, 1992). In the short term, firms find it difficult to shift strategies (Hannan and
Freeman, 1984) and may decide to refrain from phasing out a product if portfolio performance
improvements are believed to be imminent or if disconfirming information is absent.
Elevation and portfolio feedback. Elevation may amplify the persistence behavior delaying
phase-out that is characteristic of extremely over- or underperforming firms. The reason is that
slack-directed behavior (which follows performance above aspirations) and problem-directed
behavior (which follows performance below aspirations) may differ along with the level of the
firm at which phase-out decisions are made. Evidence from the behavioral theory of the firm
suggests at least two reasons for such differences. First, behavior is most strongly motivated in
areas that are considered to be important by managers and their key constituents; second,
behavior is shaped by managerial biases and the information available to them for decision
making (Cyert and March, 1963). A vertical hierarchy allocates the attention of lower and more
senior managers to narrower and broader goals, respectively (Ocasio, 1995). As a result, attitudes
toward discrepancies between aspirations and current performance—as well as the information
available to address these gaps—will vary with the structural position of decision makers.
A high-performing portfolio reflects one or more successful products, which increases the
likelihood of the firm cannibalizing its less successful or older products. At the level of product
managers, who typically manage one product after another, this concern may manifest as a desire
14
to accelerate the phase-out of certain products in advance of cannibalization. Yet social, cultural,
and economic incentives, which are reinforced by pressures from external constituents, induce
senior managers to focus primarily on the aggregate performance of the portfolio; this focus
renders them less concerned about the cannibalization of any single product. Investors and
analysts prefer continuity of strategies when performance is good (Benner, 2010), and such
pressures for consistent performance are greater at higher levels of the corporate hierarchy
because senior managers often either deal directly with the investment community or report
directly to the C-level managers who do (Kaplan and Minton, 2006; Wiersema and Zhang,
2011). If the current strategy has yielded a successful portfolio, then there is more external and
internal support for prolonging the life of products within that portfolio.
Because senior-level managers are more likely to set the overall portfolio strategy, they
are more likely to attribute any subsequent success to that strategy. In other words, when such
attributions are made by the same managers who designed the portfolio, those managers become
even more inclined to believe themselves responsible for its past favorable performance
(Milliken and Lant, 1991). Hence they are likely to implement processes that reinforce the
current portfolio mix and to be most concerned with the performance dimension that reflects
favorably on their actions. Because they are vested with better information and the capacity to
share information and deploy resources across the different product units, senior managers can
more effectively extend the life of products in reponse to constituent pressures and in support of
the status quo—provided performance is high.
The same constituent pressures for continuity following good performance are also likely
to amplify persistence among senior-level managers when portfolio performance falls below
aspirations. Investor concerns about poor performance induce managers to behave myopically
15
(Stein, 1989), which may manifest as an emphasis on short-term performance and the
elimination of spending not clearly related to short-term profitability (Cyert and March, 1963:
171). At senior levels, decisions concerning phase-out are coupled with those concerning cuts in
discretionary spending (e.g., advertising, travel, training) and human resources (e.g., by laying
off consultants or even employees), cuts that are made to increase current earnings and focus
product managers on core activities (Bushee, 1998). When performance problems arise and
portfolio decisions are elevated, higher-level managers are likely to devote fewer resources to
next-generation than to current products in attempting to boost current earnings (Bushee, 1998);
lower-level managers are less likely to exhibit this preference (Gaba and Joseph, 2012).
Senior managers may fail to recognize that a prior strategy is obsolete simply because
they do not immediately benefit from more adaptive representations of the emerging competitive
landscape, which usually originate at lower levels of the firm (Tripsas and Gavetti, 2000). Hence
these managers may not grasp fully the changing environment or may fail to seek out the source
of performance problems, and these deficiencies foster the persistence of current solutions and
products. Thus, when the portfolio is not performing well, less aggressive phase-out behavior is
likely when those decisions are made at higher than at lower levels. Consequently, we suggest
the following.
Hypothesis 3a (H3a): The elevation of portfolio decisions amplifies the extent to which
above-aspiration level portfolio performance decreases the likelihood of phase-out.
Hypothesis 3b (H3b): The elevation of portfolio decisions amplifies the extent to which
below-aspiration level portfolio performance decreases the likelihood of phase-out.
Consultation and portfolio feedback. Increased consultation may reduce the likelihood that
managers will deviate from the status quo when performance is above aspirations. In consultative
environments, the portfolio strategy and product phase-out are determined collectively, with each
16
function or unit “weighing in” on the decision. This may occur through communication channels
such as meetings to review products, planning, or design. As mentioned previously, product
groups at Motorola met collectively to discuss portfolio planning, which involved upcoming
product introductions and exits in addition to longer-term activities. These meetings included
representatives from different product groups and from various functions. When portfolio
decisions are made in this manner (i.e., with the goal of achieving consensus), excess resources
may help to resolve any potential disagreements about which products should receive further
managerial attention. Moreover, decision makers then feel collectively responsible if portfolio
performance exceeds aspirations; this leads them to interpret success as validating their
collective approach and so to favor extending the life of products in the portfolio.
However, attempting to achieve consensus among multiple interests may accelerate
phase-out when performance is below aspirations. Poor performance may deplete resources and
fragment interests, values, and beliefs; these consequences will likely intensify internal rivalries
and motivate managers to challenge the status quo (Ocasio, 1995; Birkinshaw and Lingblad,
2005). Hence the negotiation process faces less collective resistance to change—here, product
phase-out—than it otherwise would. Therefore, decision making is more likely to result in
product phase-out than when performance meets aspirations. Related research has shown that,
during periods of economic adversity, divergent interests lead to a number of organizational
changes; these include the replacement of managers (Ocasio, 1994) and the spawning of interest
groups or coalitions that might fragment organizational knowledge (Aldrich, 1999: 118).
Finally, if performance is poor then consultation may mitigate the escalation of
commitment to failing products. The divergent opinions that normally abound when consensus is
sought via collective decision making may help raise key environmental or technological issues
17
and thereby update perceptions of the firm’s positioning. Such divergence may also serve to
highlight any mis-specified attributions for performance. When performance is poor, external
attributions dominate and this reduces the unit’s responses to feedback. However, if portfolio
management decisions must be approved by various units and performance is poor, then the
beliefs underlying those attributions are more likely to be questioned. These considerations lead
to our final hypotheses.
Hypothesis 4a (H4a): Increased consultation in portfolio decisions amplifies the extent
to which above-aspirations portfolio performance reduces the likelihood of phase-out.
Hypothesis 4b (H4b): Increased consultation in portfolio decisions attenuates the extent
to which below-aspirations portfolio performance reduces the likelihood of phase-out.
METHODS
Our study looks at product exit and organizational structure in the mobile device industry during
the period 2004–2009. This is an ideal setting to examine their relation because the industry
leaders constitute a relatively stable set and because most mobile devices have a short product
life. Since there is no single source that combines device-level information with an adequate
description of how phase-out decisions are made, in this study we employ qualitative measures
of organizational structure in addition to detailed, device-level quantitative data. These data were
obtained from a variety of sources, including public filings, press and literary coverage, internal
records, and interviews that concentrate on the five largest participants (in terms of units sold) in
the mobile device industry. These companies accounted for nearly two thirds of all mobile
devices launched on the German market during the time period of our study; they also were the
industry leaders, during this period, in establishing the pace of product turnover and development
as well as the major device strategies in the marketplace.
18
The quantitative data sample covers all mobile devices from our five focal firms that
were introduced to the German market after January 2004 and discontinued before December
2009. Since a typical phone lifetime is only 4.3 quarters, this period captures multiple updates of
handset manufacturers’ product portfolios during a time that saw significant changes in the
industry, including development of 3G (third-generation) data standards. This period is
considered the era of feature phones, and is prior to the move toward smartphones. The GfK
global retail panel served as our primary source of performance data. This panel is regarded as
the industry benchmark in data collection because it gathers retail sales figures at the point of
phone sales to consumers rather than from manufacturer surveys. The GfK data are also unique
in providing phone-level price and sales information. We supplemented and cross-checked the
GfK data set with data from competing providers, such as the World Cellular Information
Service, Informs World Cellular Handset Tracker, technical websites such as gsmarena.com, and
the Strategy Analytics online database.
As mentioned previously, the mobile phone industry is an ideal setting to examine the
effects of performance feedback on new product introductions. First, it is a high-velocity
industry characterized by a high rate of new product introductions and technological advances
(Eisenhardt, 1989, Keil, McGrath, and Tukiainen, 2009). Second, the major players did not
change during the period under study even though their relative performance varied. Third, the
critical goals of firms in the industry—for instance, unit sales and replacement pipeline—are
widely shared and watched by competitor firms and industry analysts.
To ensure accuracy, a subsample of the data was selected and checked against publicly
available data on product availability and features. The GfK revenue figures were also checked
against Datamonitor estimates, and the former were found to be in line with the latter. Data were
19
analyzed by quarter; the result was a total of 1,307 product-quarter observations, comprising 337
product exits across 361 devices within the sample. Firm financial data was acquired through
Compustat and quarterly reports. Descriptive statistics and correlations for all predictor variables
are given in table 1.
*** Insert Table 1 about here ***
Dependent Variable
We explore the effect of different decision-making structures and performance–aspiration gaps
on product phase-out or discontinuation. In the mobile device industry, the product life cycle is
approximately four quarters; see the histogram presented as figure 1.
*** Insert Figure 1 about here ***
The key event for our analysis is discontinuation of a mobile device. Unfortunately, the
precise date of a product line’s discontinuation is difficult to establish (de Figueiredo and Kyle,
2006) because such dates are seldom known outside the company. Mobile device phase-outs are
also complicated by retail distribution channels, which purchase phones from the manufacturers
before selling them to end users. The inventory in these retail channels results in sales appearing
to continue long after phones have been discontinued by the manufacturer. We therefore define
discontinuation as the cessation of product shipments from the handset manufacturer, not as the
cessation of retail sales. The data give an indication of manufacturers’ internal selection
decisions in the form of a sudden and discontinuous fall in monthly sales, although they may not
fall to zero. For a small subsample of devices we were able to locate data through internal
company documentation that provided the exact discontinuation date.
All device phase-outs were coded manually by two individuals. Accuracy across
individuals was high: for 94 percent of the devices, the coded discontinuation dates were within
20
three months of each other. The coding was also checked against available internal data,
revealing near-perfect accuracy. For those few devices whose coded discontinuation dates were
not within three months, the coders jointly revisited the available information and reached a
consensus on the discontinuation date. The coders did not identify phase-outs of any devices for
which the observations were either left-censored (i.e., that were already in the sample at t = 0,
which would exclude them from the analysis) or right-censored (i.e., that may or may not have
been discontinued by the end of the observation period).
Organizational Structure and Decision Making
Following an approach similar to that of Henderson and Cockburn (1994), we conducted a
literature search to build an initial understanding of each firm’s organizational structure and
decision-making processes. This search covered more than a dozen newspapers and magazines,
several business case publishers, and books spanning a decade of coverage on our five focal
companies: LG, Motorola, Nokia, Samsung, and Sony Ericsson. We conducted semi-structured
interviews with individuals knowledgeable about the product management process in order to
develop a narrative of key structural characteristics of each firm across our observation period.
These interviews were approximately an hour in length; they were held in person when possible
and by telephone otherwise. Interviewees were employees of the firm who were familiar with the
intricacies of the product management process—by virtue of playing a product management role
themselves or of interacting extensively with the organization’s portfolio management teams.
We used these initial interviews and narratives to develop a questionnaire that could be
used to measure some relevant structural aspects of decision making that were common across
all firms. This questionnaire specifically prompted the respondent to answer each question for
each year of the study period. We developed an assessment of each firm that was based on our
21
narratives and then used this assessment to test the accuracy of our secondary data by sending the
questionnaire to previous interview respondents at each firm for verification purposes. To
measure elevation, we asked respondents questions about how often portfolio management
decisions were deferred to higher-ups in the organizational structure. To measure consultation,
we asked respondents whether or not portfolio management was driven by participative
consensus. For each year of the study period, responses are given on a 5-point Likert scale on
which 1 (resp., 5) represents the least (resp., most) elevation within the hierarchy or consultation
among same-level units. The survey instrument was translated into Korean in order to facilitate
interviews with Korean firms (Harkness, Pennell, and Schoua-Glusberg, 2004).
Rather than embed our two measures in a larger, “organizational structure” construct, we
chose to analyze separately the effect of each structural aspect on organizational decision
making. One advantage of this approach is that it parallels the theoretical difference between
centralization and integration explored previously; another advantage is that is yields greater
insight into the effect of each structural aspect on culling. Moreover, separating consultation
from the locus of decision-making authority is consistent with our experience in assembling the
qualitative data, which indicated that higher-level staff involvement in managing products varies
independently of the firm’s reliance upon consultation in decision making.
Aspiration levels and performance feedback. Feedback consists of performance signals
derived from multiple sources—in particular, one’s own historic performance and the
performance of competitors within a competitive space. These sources of feedback have been
classified by empirical studies as, respectively, historical and social aspiration levels (Greve,
1999). Consistent with other research on organizational learning in the context of a product
portfolio, we focused initially on historical aspirations at the portfolio level (Audia and Greve,
22
2006). Less is known about how managers form reference groups (Greve, 1998); hence defining
the social group relevant to the devices against which a manager makes comparisons would
prove difficult in the mobile device industry, where devices compete—with respect to features
and prices—not only within a given product generation but also across generations.
Our performance measure is based on unit sales at the portfolio level because such sales
are a key performance indicator in the mobile device industry and are tracked closely by
manufacturers, carriers, and analysts. Although we control for the sales performance of
individual devices, our main independent variable is portfolio sales. Theoretically we are
interested in portfolio feedback. During this time period, only a handful of devices (e.g., the
Motorola RAZR) garnered managerial attention in excess of normal portfolio management
procedures. Furthermore, “hero” devices (e.g., the Samsung Galaxy) had not yet arisen in the
portfolios to which this study pertains. Portfolio unit sales are simply the total sales of all of a
firm’s devices within a given period. To account for life-cycle issues in the portfolio that affect
managers’ perceptions of performance relative to aspirations, we use the slope between portfolio
unit sales at time t and at time t − 1 as a measure of the percentage change in those sales. This
change is our measure of performance, which is consistent with the approach taken by Audia and
Brion (2007).
From a conceptual standpoint we are arguing that managers are concerned less about
whether or not portfolio sales increase or decrease relative to past levels than about whether or
not those sales has achieved management’s growth targets. Thus formulating aspiration levels in
terms of growth accords with our product manager interviews, with internal presentations used to
guide culling decisions, and with previous research (e.g., Audia and Brion, 2007) on aspirations
and performance. As a robustness check we also used percentage change in portfolio total
23
revenue. Our results were similar under both measures, which suggests that pricing changes
before discontinuation do not fundamentally alter the feedback processes examined here.
We employ a formulation similar to those used in previous studies on performance
feedback (Greve 2003b; Audia and Greve, 2006) to define a historical aspiration level for
product portfolio performance as follows. Let HAuit denote the historical aspirations of firm i at
time t at the uth level, where u represents portfolio share, and let Puit denote the performance of
that firm. Then historical aspiration is given by HAuit+1 = αPuit + (1 − α)HAuit, where the
adjustment parameter α is chosen by searching over all parameter values—in increments of
0.1—and then using the value that produces the maximum log-likelihood (i.e., 0.8) consistent
with previous studies (Cyert and March, 1963; Greve, 2003a). Our findings are generally robust
to the weighting scheme.
The performance–aspiration gap is simply defined as the difference between performance
and aspiration level for each of the portfolios. We implemented a spline function to compare the
effects of this gap above and below the aspiration level (Greve, 2003b; Miller and Chen, 2004).
This was accomplished by splitting each of the historical performance variables into two
variables. Performance above aspiration is set equal to zero for all observations in which the
performance (at the portfolio or phone level) of the focal firm is less than its historical aspiration
level, and it equals the difference between actual performance and the historical aspiration level
when the firm’s performance is above that level. Thus,
[Performance above historic aspirations](u )  max[0, ( Pitu  H itu )],
where u represents one firm’s product portfolio. Performance below historic aspiration is defined
symmetrically. In other words, it is set equal to zero when performance is above the aspiration
level and equals the performance–aspiration gap when performance falls below that level:
24
[Performance below historic aspirations](u )  min[0, ( Pitu  H itu )].
All gap variables are lagged by three quarters in the empirical specification; note that
Henderson and Stern (2004) also lagged key independent variables in their event history analysis
of product culling behavior. Weighing the relatively frequent re-evaluation of performance
results against the constraints on device life and the ramp-down time required to coordinate all
aspects of the phase-out process, we decided that a lag of three quarters would be appropriate for
illustrating the relationship between structure, feedback, and phase-out while maintaining
acceptable temporal separation between independent and dependent variables. Given the mean
product life for a mobile device consists of only 4.3 quarters on average, a longer lag structure,
such as four quarters, is inconsistent with both internal review cycles and our semi-structured
interviews with individuals involved in product management.
Device sales and replacement as important factors in phase-out. In addition to the key
theoretical variables described already, two of our controls warrant special attention. First, we
control for phone sales performance to account for the impact (if any) of a particular device’s
sales on the choice of product(s) to be phased out of a firm’s portfolio. Thus, the interaction
effects we observe between structure and feedback are beyond those associated with the sales
performance of any individual device. Although this variable is included in the model as a simple
control for device performance, we also tested its robustness to alternative formulations—
specifically, as a feedback variable—that were consistent with observed main portfolio feedback
effects.
Second, we include a control for whether or not the device has a direct replacement. Such
devices can be quickly pulled from the market, regardless of performance, to make room for the
replacement device. All else equal, this may affect how the phase-out decision is made and
25
create a constraint that transcends feedback processes. For our purposes, a direct replacement is
one that is launched within one quarter of the previous device’s exit and that has a launch price
within 10 percent of its predecessor device’s launch price. Our definition is consistent with
pricing patterns observed in the industry, where devices are quickly discounted after launch and
marketers tend to “tier” phone offerings into price bands. In practice, this is a fairly strict
standard for what constitutes a replacement. We tested other formulations of this variable in
which the window between exit and launch was lengthened to two quarters in the presence of
various price differences. Our results were not affected by these alternate constructions.
Other Controls
Our other selected controls provide insight into alternative mechanisms that might be active in
the phase-out decision process. For this study, these additional controls were motivated by
interviews with product managers at the major firms, internal documentation related to culling
decisions, and the relevant theoretical literature on firm decision making. We use firm sales as a
proxy for size, whose effect on product exit decisions we wish to assess (cf. Hannan and
Freeman, 1984; Henderson and Stern, 2004). In estimating the model, we control for a number of
variables in order to isolate their impact on product discontinuation; these variables capture
characteristics of the individual device, its parent firm, and the market as a whole for each period
of time. Our model also includes year and firm dummies to test for robustness (see table 4).
Several firm-specific variables were also used to control for heterogeneity among
participants in the marketplace—specifically, with respect to the number of 3G devices and
smartphones in a firm’s portfolio—so as to capture broad improvement in underlying
technologies (Greenstein and Wade, 1998; de Figueiredo and Kyle, 2006). We used a firm’s
average portfolio age at each moment in time to capture the degree of portfolio obsolescence
26
(Henderson and Stern, 2004), and we used cash levels and current ratios (i.e., assets/liabilities) to
account for different levels of slack across firms (Cyert and March, 1963).
Previous studies have employed unit counts over time to approximate experiential
learning (de Figueiredo and Kyle, 2001; Henderson and Stern, 2004). We take an analogous
approach when measuring capability development over time, using total number of phones
launched since the start of the observation period as a metric for phone introduction capabilities.
We also included measures of culling experience that are similar to those used by Sorenson
(2000). Our regressions incorporated an indicator for whether or not the firm manufactures its
devices in-house as well as one for whether or not it owns a semiconductor unit. These controls
were meant to capture the presence (or absence) of those firm capabilities and to account for
differing levels of vertical integration across firms; this is important because upstream
capabilities may affect behavior in downstream markets (de Figueiredo and Teece, 1996).
Finally, we controlled for both market density and the number of same-period competitive phone
launches (after Sorenson, 2000; Henderson and Stern, 2004) to account for period-specific
market dynamics that may drive decision making and for the extent of market crowdedness. All
time-varying controls were lagged three periods so that they would be consistent with our lag
structure for performance aspiration variables.
Empirical Specification
In order to examine the factors that affect phone life over time, we apply a piecewise exponential
hazard rate model with many similarities to the approach of Sorenson and Stuart (2000) and
de Figueiredo and Kyle (2001, 2006). This model accounts for right-censoring—although leftcensored observations, for which we lack data on the beginning of a phone’s life, are dropped
from the sample—and it offers flexibility in handling both time-invariant and time-varying
27
covariates. Whereas the exponential specification assumes a constant and time-invariant hazard
rate, the piecewise specification enables us to apply different base hazard rates that depend on
the device’s age; thus we can control for any heterogeneity in decision-making processes that is
driven by age-dependent factors. The coefficients in the estimation can be viewed as generating
multipliers of the appropriate underlying base hazard rate, which increases as the mobile device
ages. The clock in this model is device age, and the individual device is the unit of analysis. In
the piecewise analysis, our pieces are the following intervals of device age: 0–1 quarters, 1–2
quarters, 2–3 quarters, 3–4 quarters, and more than 4 quarters. A Kaplan–Meier survival graph
(see figure 2) indicates that the hazard rate differs for each of these time periods, so the intervals
we use lead to legitimate improvements in model fit.
*** Insert Figure 2 about here ***
Our specification follows that in Blossfeld, Golsch, and Rohwer (2007), and we assume a
constant hazard rate for each interval of device age: r (t )  a . The underlying survivor function
within a piece is G (t )  e  at and the hazard rate can be expressed as ai  exp{xi  } , which
implies different hazard levels for different xi observations. The model’s beta coefficients are
estimated using maximum likelihood techniques.
RESULTS
Table 2 shows the piecewise exponential hazard rate results for quarterly device phase-out.
Model 1 includes only the control variables, models 2 and 3 examine (respectively) the impact of
organizational structure variables related to elevation and consultation, model 4 examines both
impacts jointly, model 5 shows the effects of performance feedback, models 6 and 7 show the
interaction between performance feedback and the organizational structure variables, and model
8 shows all main effects and interactions together. The baseline model exhibits a highly
28
significant fit to the data (a chi-square test reveals p < .001). The inclusion of both structural
characteristics yields significant improvement in model fit for models 2 and 3, as indicated by
the significant coefficient for these variables. In model 4 the change in χ2 over the base model 1
is 5.48, which is below a p < .01 threshold for increase in model fit. This finding suggests that
the two structure elements jointly yield a significantly better fit to our data than does either
element independently. To interpret the magnitude of the coefficients appearing in table 2, we
calculated the effect that a one standard deviation change on our measures of elevation and
consensus would have on the instantaneous hazard rate based on the main effects reported in
table 2, model 4. In addition to being statistically significant, the practical significance of a
change in structure is quite large, the instantaneous hazard rate changing by 28 and 35 percent in
response to a one standard deviation change on our measures of elevation and consensus,
respectively. Moreover, the models that include interactions between structure and feedback
exhibit a noticeable improvement in fit over baseline models that incorporate performance
feedback only (for models 6, 7, and 8; p < .001). Figure 3 summarizes our hypotheses and
results.
*** Insert Table 2 about here ***
*** Insert Figure 3 about here ***
In model 1, we find that competitive density drives an increased rate of product turnover
and that the presence of a replacement device leads to a quicker rate of product exit. These
results are consistent with our qualitative interviews, which suggested that managers actively
engage in scanning the competitive landscape and juggling different generations of products. We
find highly significant results on device-level performance, confirming the intuition that culling
is strongly influenced by the unit sales of individual devices.
29
We also find a significant effect on size (measured in terms of corporate sales), which is
associated with reduced product culling. This result echoes the literature addressing the effect of
organizational inertia on adaptation and innovation (Hannan and Freeman, 1984), as the
organizational inertia associated with increased size has the effect of slowing down the rate of
product culling. Contrary to results in Sorenson (2000) and Henderson and Stern (2004), the
coefficient for our “experience” variable is not significant; in other words, current culling rates
seem largely unrelated to previous rates in the same market. There are numerous reports of the
difficulties that traditional mobile device firms encountered when adjusting to the demands of
new technologies such as 3G (for a description of Nokia’s struggles in this area, see Troianovski
and Grundberg, 2012). Note also that the rate of product turnover (4.3 mean quarters of life) was
much higher in our study than in that of Sorenson (2000), who examined workstations (2.84
mean years of life), and Henderson and Stern (2004), who examined personal computers (2.31
mean years). These considerations explain why an experience variable, which captures learning
over a long period of time, may be less relevant in our context than in previous studies.
It is interesting that these two technological developments had the opposite effect on
hazard rates. Whereas increasing the number of 3G devices in a product portfolio reduced hazard
over time, additional smartphones were associated with an increased hazard rate. This result may
reflect key differences in these technologies: acceptance of the 3G standard was instrumental in
the widespread use of smartphones, but no dominant designs emerged for such devices until after
the sample period. Consequently, handset manufacturers may have been forced to cull earlylaunch smartphones at a higher rate while consumer preferences were evolving.
Model 2 adds the elevation measure to our culling model. Hypothesis 1 argued that
elevation will generally speed culling because higher-level managers can more quickly transfer
30
resources from one product to another. Consistent with H1, we find a statistically significant
coefficient for elevation in model 2; that coefficient is significant and positive excepting only in
models 4 and 7, which indicates that higher elevation is associated with increased culling. Even
in models 4 and 7, the effect of elevation on phase-out remains significant at the 10 percent level.
Model 3 shows the main effect of consultation on the culling model. Hypothesis 2 argued
that, when consensus is required, more bargaining occurs and more trade-offs are made; these
factors should result in products being left on the market longer, on average, despite product
managers’ efforts to move on to the next generation. We thus expect the sign on the consultation
variable’s coefficient to be negative. Our results corroborate this hypothesis: in model 3 we find
a statistically significant negative coefficient for consultation, indicating that increased
consultation is associated with slower phase-out. This finding is robust and significant (p < .01)
in models 4, 6, 7, and 8.
Model 5 shows how our measures of portfolio-level performance feedback affect product
phase-out. Consistent with H3a and H3b, we find that performance both above and below
aspirations is associated with a slower rate of product phase-out. This finding accords with our
argument that product management faces pressures driven by the continued commitment to
previously made decisions and by both internal and external attributions of product success or
failure.
Models 6, 7, and 8 show the results when structure interacts with feedback. Model 6
shows feedback being moderated by elevation, model 7 shows it moderated by consultation, and
model 8 combines both of these variables. Following Sorenson (2000), we simultaneously add
each term and its interaction to our models. It is important to note that the “main effect” of
feedback cannot be interpreted as simply the coefficient for the portfolio-level feedback variable
31
(Jaccard and Turrisi, 2003). As these authors stated, the coefficient for the main effect represents
its influence when the other term in the interaction is zero. In our sample, the zero value is
meaningless, as it does not occur, as consultation and elevation vary from 1 to 5. We find a
negative effect of portfolio feedback on culling when performance is above aspirations. Models
6–8 exhibit similar main effects of the structural variables as in models 2 and 3, and—for
performance above aspirations—each effect in the interaction models is preserved when both
sets of interactions are run simultaneously.
The formal tests for hypotheses 3a, 3b, 4a, 4b, 5a, and 5b amount to examining the
regression output (table 2) and comparing it against our expected signs for these variables
(figure 3). Models 6–8 corroborate our hypotheses, though with only partial support for H4b.
Observe that the model with both sets of interactions is the one that best fits the data when we
compare models 2 and 3 to models 6–8; this is our direct comparison of the “main effect” models
with the “moderating factor” models that include portfolio feedback effects (the chi-square
statistic for this comparison is significant at p <. 001). We take the outcome of this test to mean
that structural characteristics explain culling better when considered jointly than individually.
To facilitate the interpretation of these findings, we have separately graphed the key
results of models 6 and 7 in figures 4 and 5, respectively. The x-axis of these figures marks
performance relative to aspirations, and the two are equal at the center of this axis. In each
figure, the y-axis represents the (log of the) multiplier of the hazard rate of phase-out (cf. Davis
and Greve, 1997); when this value is greater (less) than 1, the likelihood of phase-out increases
(decreases). Two lines representing different values of elevation and consultation are plotted in
this graph of figures 4 and 5, respectively. Thus the graphs show how elevation and consultation
affect the rate of phase-out at different levels of performance relative to aspirations. The domain
32
of each graph is approximately ±1 standard deviation from the mean for performance relative to
aspirations suggesting that the entire domain is relevant for analysis. The chosen values of
elevation and consultation are also within the data range.The interaction terms for both elevation
and consultation are significant in models 6 and 7: all of the displayed slopes are statistically
different from zero. However, Figure 5 shows that, at certain points in the structure–performance
space, the multiplier of the hazard rate is 1 (i.e., no net effect).
*** Insert Figure 4 about here ***
Figure 4, which plots elevation against performance relative to aspirations, is the source
of several striking results. It is most interesting that the response to feedback depends on where
decision rights reside in the organization. The net effect of performance feedback can be either to
hasten or delay phase-out, which suggests an interesting link between the behavioral mechanism
of performance feedback and the structure of a firm’s decision making. In particular, the right
hand side of figure 4 shows a range of performance less aspirations in which a lower degree of
elevation is associated with a higher rate of product phase-out.
Three other observations are warranted by examining figures 4 and 5. First, the multiplier
of the hazard rate is lower at higher levels of performance relative to aspirations. This dynamic is
consistent with hypotheses 3a and 3b. In the context of figure 4, our hypotheses 4a and 4b
suggest that the hazard rate multiplier declines the most when elevation is highest. A comparison
of the slopes of both level curves of elevation clearly demonstrates this. When performance
equals aspirations, the multiplier is larger when elevation is higher. Yet at moderate levels of
performance above aspiration, the multiplier is smaller for higher than for lower elevation; this is
the effect of the plot for performance above aspirations being steeper when elevation is high,
consistent with H4a. A similar pattern is present with respect to H4b when we compare the slope
33
of the lines where elevation = 5 to elevation = 3 on the left-most graph; here the effect is less
dramatic, however, and is insignificant at the 5% level in the full model.
*** Insert Figure 5 about here ***
Figure 5, which addresses consultation, appears to be less variable than its elevation
counterpart (figure 4), though the regression output suggests that both interactions between
performance feedback (i.e., either above or below aspirations) and consultation are significantly
different from zero. The range of figure 5 is strictly less than 1; this suggests that, in the
displayed performance feedback–consultation domain, the net effect of these variables is to slow
phase-out. Consistently with H2, at higher levels of consultation we find a lower multiplier of the
hazard rate: both lines where consultation = 5 are lower than the lines where consultation = 3 for
the range of the data.
Altogether these results suggest that, under different conditions of portfolio growth, the
configuration of the organization with the greatest net increase or decrease in culling might be
drastically different. In particular, our results indicate that phase-out is most rapid in firms that
outperform aspirations and for whom decision rights are relatively decentralized to lower-level
managers. However, the opposite holds when portfolio performance is below aspirations:
elevation of decisions to those higher in the hierarchy is associated with increased culling.
Sensitivity Analysis
In this section we explore the sensitivity of our findings to two alternative specifications: one in
which the independent variable for performance feedback is formulated using revenue figures;
and one in which firm and year dummy variables are included, as is common in product-level
analyses.
34
Alternative formulation of performance feedback variables. In order to assess the robustness
of our results to alternative explanations, we have included a number of robustness and
sensitivity tests to ensure that the observed patterns involving organizational architecture and
performance feedback are consistent across specifications. First, performance feedback is usually
measured with respect to more than one organizational goal (Cyert and March, 1963). We have
therefore recalculated our results, which were obtained from sales figures, while using revenue
figures instead. Table 3 shows the results when models 1–7 are recalculated based on revenue.
*** Insert Table 3 about here ***
Overall, the results are consistent with those reported in table 2. This suggests that pricing
strategies do not significantly alter the nature of the feedback process that occurs at the portfolio
level of aggregation. We believe that the results determined from sales figures are more reliable;
the reason is that revenue figures must be based upon a single price estimate because our data are
not such that we can disaggregate the pricing strategy of individual retailers. Hence one should
expect more measurement error in the revenue variables, which would bias our findings toward
non-significance. The revenue approach yields smaller log-likelihood figures, which indicates
that models based on unit sales better fit the data. This explains why we have chosen to
emphasize results based on the latter approach, though we remain encouraged by the fairly
consistent results across the different formulations.
Year and firm dummies. Next we incorporate both firm and year dummies to account for
organizational and temporal idiosyncrasies—for instance, time-invariant firm characteristics or
accelerating technological advancement—that may broadly affect phase-out patterns. Both year
and firm dummies are commonly used in models of product management (de Figueiredo and
Kyle 2001). Table 4 shows the results of adding these effects sequentially: first the firm dummy
35
and then the year dummy. These models also parse out the effect of organizational structure on
measurement of the key variables, demonstrating that our main results hold even when we
control for cross-firm differences in measurement.
*** Insert Table 4 about here ***
When these dummies are included in the regression, the results are virtually identical to
those reported in table 2, model 8. The interaction between consultation and performance above
aspirations is no longer significant at the p < .05 level, but with year and firm fixed effects the pvalue is .052 and remains significant at the 10 percent level. Note that incorporating firm and
year dummies does not improve the model’s statistical fit. It is therefore reasonable to conclude
that a model excluding these parameters yields a statistically similar and thus adequate fit to the
data, which supports our decision to omit these effects from the model’s original formulation.
DISCUSSION
This study develops a model of situated selection to explain the effects of organizational
structure on persistence in product phase-out. In particular, we examine the effects of decisionmaking elevation and consultation on product phase-out in the mobile device industry. Our
results suggest that decision makers at lower levels of the hierarchy are constrained in their
ability to coordinate the various activities required for phase-out and thus in their tendency to
cull products more slowly. In contrast, when phase-out decisions are made at higher levels, the
decision makers have better information, are less encumbered by parochial agendas, and are
better situated to recognized interdependencies among product teams and to support products
that succeed in the marketplace. Yet higher-level managers are also more prone to inertial forces
and attributional biases and tend to favor the status quo. This is because their attention is focused
at the portfolio level, not the product level, an orientation that is driven in part by the demands of
36
external constituents and their separation from product market activities. As a result, we observe
persistence in the product portfolio when its performance is especially high or low.
We also find that, in a structure under which decision-makers engage in relatively less
consultation with other managers, product life is typically shorter because bargaining is less
prevalent and negotiated trade-offs are less likely to occur. In such cases, managers are required
neither to secure others’ approval for product proposals nor to accommodate the often divergent
interests of the managers responsible for competing products. In contrast, consensus-driven
environments delay phase-out when portfolio performance is above aspirations because that
success is attributed to the collectively established decisions. Yet when performance is below
aspirations, a consultative environment may encourage more debate and a willingness to
question prevailing explanations of that underperformance, resulting in faster phase-out.
Our study makes several contributions. First, we offer new insights into the determinants
of product phase-out. Much of the research on product exit points to drivers that are economic
(Greenstein and Wade, 1998) or based on prior experience (Henderson and Stern, 2004). We add
to this research by highlighting the importance of organizational structure. In particular, our
emphasis on its role in situating phase-out decisions at lower levels—and in absence of other
managerial input—presumes that selection is, in practice, a more local determination than
previously considered. Implicit in much of the seminal research (e.g. Burgelman, 1991) is that
internal selection environments are relatively homogeneous, so these studies largely ignore
variations in where and how decisions are made. Our study of leading mobile device firms
reveals that the locus of decision making varies, and we posit that selection patterns may vary as
a result. Selection decisions do not always reside at the lowest levels; in some cases, they reside
quite high in the hierarchy. At Motorola, for example, in 2009 the CEO and head of the mobile
37
devices unit sought to clear out the entire portfolio in attempting to help save the division. We
offer a model of situated selection that accommodates the diversity, within and across firms, in
decision procedures.
Our study also underscores the distinction between hierarchical and consultative decision
making. Much of the research on centralization and decentralization implicitly equates the mere
presence of decentralization with greater interactions between managers; however, we do not
make this assumption. In fact, we discover that product phase-out is delayed unless portfolio
decisions are made at the firm’s lower levels and middle managers have much discretion. Thus
our findings highlight the key role of middle managers (cf. Huy, 2002), and support research that
suggests the attention-directing qualities of structural differentiation (but not integration) may
increase adaptability (Ethiraj and Levinthal, 2004; Joseph and Ocasio, 2012).
Second, this study links behavioral theories of performance feedback to organizational
structure as a means to explain the complexity of phase-out decisions and the implications for
persistence behavior following perceptions of success and failure. In doing so, we have
augmented the growing scholarship on performance feedback by considering some important
conditional effects imposed by elevation and consultation in decision making. The role of
organizational structure has been notably absent in studies of performance feedback (Gaba and
Joseph, 2012), and our study suggests that more research linking cognitive and structural
explanations for adaptive behavior is in order. Moreover, it is clear from our results that the
forces inducing persistence are not dispersed uniformly throughout the firm, and more work
exploring such differences could be illuminating. In a related vein, our study also augments
models of shared cognition that focus on the performance implications of broadly diffused
mental models, schemas, frames, and logics (Eggers and Kaplan, 2009; 2013). Here we suggest
38
that capabilities may be created not through shared cognition but though situated cognition: the
spatial and temporal conjunction of certain players and types of feedback.
By extension, this paper documents that the role of organizational structure in decision
making involves more than information processing. In particular, we establish the effect of
structure on situated decision making and demonstrate that responses to performance feedback
vary with the elevation of and consensus sought in decision making. Organizational structure
situates cognition and hence variations in perceptions, interpretations, and responses to
performance feedback. Our study is consistent with recent research suggesting that cognitive
biases may be affected, and in some cases circumvented, by the organizational context in which
learning and decision making occur (Lave and Wenger, 1991; Elsbach, Barr, and Hargadon,
2005). These studies argue that contextual dimensions—such as organizational culture
(Edmondson 1999, Bunderson and Sutcliff, 2003), social identity (Kane, Argote, and Levine,
2005), and organizational processes (McNamara and Bromiley, 1997)—interact with cognitive
factors to situate learning processes and alter outcomes (Argote and Todorova, 2007). For
instance, McNamara and Bromiley found that both organizational and cognitive factors influence
risky decision making but that, when both are present, organizational factors (e.g., goals) tend to
dominate cognitive biases. This means that models of decision making and cognition may be
inaccurate in their predictions of real-life processes, such as internal selection, if they fail to
consider such important situational factors (Lant, 2002) as the decision-making structure of
complex organizations. The new insights we provide on this score suggesting that the importance
of structure may lie not only in its capacity for responding to environmental complexity,
speeding decisions, and processing information—as documented in prior research—but also in
its ability to shape the context in which portfolio decisions are made.
39
This also points to one of the limitations of the study in that we examine a very specific
period in the mobile device industry. This was a period of great turbulence and technological
change since no dominant design had yet emerged and 3G technology was in its infancy. The
iPhone was launched toward the end of our sample period and the network externalities provided
by the Android and iOS operating systems had yet to emerge. In addition, the global mobile
device industry has relatively few large players; it may be that a fragmented industry with many
small players characterized by flat hierarchies exhibits no significant effects for the level of
decision making. Second, because top management focuses on a host of goals, future research
should consider the consequences of attention to multiple simultaneous goals at both the
corporate and business unit level. It may also be that the degree to which goals are
interdependent has measurable consequences (Ethiraj and Levinthal, 2009). Highly correlated
goals may lead to different outcomes than goals derived from performance measures that are
either uncorrelated or negatively correlated. Third, we studied only the five major participants in
this industry because we lacked accurate historical structural data for the sample firms that were
less active in the mobile device arena. Future research might usefully investigate whether the
level of decision making and the extent of consultation during that process account for
meaningful variance in smaller firms, too. Fourth, our study controls for phone replacement but
captures the impact of architecture on only one aspect of the product management cycle: the
decision to phase out a product. Applying an event history methodology to a study of entry is
complicated by the difficulties in establishing when a particular observation enters the risk set
(Nerkar and Paruchuri, 2005). Neither of the conventional solutions proposed to address this
issue yields an adequate resolution. A Poisson model of entry by period is a less intuitive match
to the rate of decision making and so introduces many other, unobservable decision factors; and
40
establishing entry into the risk set as a function of the time delay between first and second
iterations of a product type is untenable given the nature of technical progress in this industry.
New ideas and work on this problem would be welcome.
CONCLUSION
From the managerial perspective, our model of decentralized autonomy may be the most suitable
approach if rapid phase-out is required. However, managers under these conditions may miss the
full life cycle of products, prematurely phasing out products that could still contribute to the
firm’s bottom line. It is worth noting that Apple, when controlled by Steve Jobs, was quite
successful in a decision-making environment that was centralized and not based on consensus.
However, the business model of Apple was much different from that of the large manufacturers
(e.g., Nokia and Samsung) and led to very few product introductions, which made phase-out less
of a concern.
For scholars, a broader contribution of this paper is to link key pillars of the Carnegie
School: hierarchy, aspirations and cooperation. These three pillars have been largely developed
independently, yielding a wealth of theory for each that is largely devoid of the other two. This
study integrates them in support of Neo-Carnegie scholars who call for a focus on “renewed
behaviorally plausible, decision-centered perspective on organization” (Gavetti, Levinthal and
Ocasio, 2007: 525). To that end, we offer a situated decision-making theory of organizational
choice, which integrates hierarchy, aspirations and cooperation amidst conflicting interests, and
provides a sharper relief of persistence in organizations.
41
REFERENCES
Aldrich, H. E.
1999 “Organizations evolving.” Sage Publications Limited.
Alonso, R., W. Dessein, and N. Matouschek.
2008 “When does coordination require centralization?” The American Economic Review,
98(1): 145-179.
Argote, L., and P. Ingram.
2000 “Knowledge transfer: A basis for competitive advantage in firms.” Organizational
behavior and human decision processes, 82(1): 150-169.
Argote, L., and G. Todorova.
2008 ”Organizational learning.” International Review of Industrial and Organizational
Psychology, 22: 193-234.
Audia, P. G. and S. Brion.
2007 “Reluctant to change: Self-enhancing responses to diverging performance measures.”
Organizational Behavior and Human Decision Processes, 102: 255–269.
Audia, P.G., and H. R. Greve.
2006 “Less likely to fail: Low performance, firm size, and factory expansion in the
shipbuilding industry.” Management Science, 52: 83-94.
Audia, P. G., E. A. Locke, and K. G. Smith.
2000 “The paradox of success: An archival and a laboratory study of strategic persistence
following radical environmental change.” Academy of Management Journal, 43(5): 837853.
Audia, P., and O. Sorenson.
2001 “Aspirations, performance and organizational change: a multi-level analysis.” Working
paper, London Business School.
Baumard, P., and W. H. Starbuck.
2005 “Learning from failures: Why it may not happen.” Long Range Planning, 38(3): 281-298.
Banbury, C. M., and W. Mitchell.
1995 “The effect of introducing important incremental innovations on market share and
business survival.” Strategic Management Journal 16:161-182.
Benner, M. J.
42
2010 “Securities analysts and incumbent response to radical technological change: Evidence
from digital photography and internet telephony.” Organization Science, 21(1): 42-62.
Birkinshaw, J., and M. Lingblad.
2005 “Intrafirm Competition and Charter Evolution in the Multibusiness Firm.” Organization
Science 16(6): 674-686.
Blossfeld, H., Golsch, K. Rohwer, G.
2007 “Event History Analysis with Stata.” Psychology Press, London, UK.
Brockner, J.
1992 “The escalation of commitment to a failing course of action: Toward theoretical
progress.” Academy of Management Review: 39-61.
Bunderson, J. S., and K. M. Sutcliffe.
2003 “Management team learning orientation and business unit performance.” Journal of
Applied Psychology, 88(3), 552.
Burgelman, R.A.
1984 “Designs for corporate entrepreneurship in established firms.” California Management
Review, 3: 154-166.
Burgelman, R.A.
1991 ”Intraorganizational Ecology of Strategy Making and Organizational Adaptation: Theory
and Field Research.” Organization Science. 2(3): 239-262.
Burgelman, R.A.
1994 ”Fading Memories: A Process Theory of Strategic Business Exit in Dynamic
Environments.” Administrative Science Quarterly, 39(1): 24-56.
Burgelman, R. A.
2002 “Strategy as vector and the inertia of coevolutionary lock-in.” Administrative Science
Quarterly, 47(2): 325-357.
Bushee, B. J.
1998 “The influence of institutional investors on myopic R&D investment behavior.”
Accounting review: 305-333.
Cannon, M. D., and A. C. Edmondson
2001 “Confronting failure: Antecedents and consequences of shared beliefs about failure in
organizational work groups.” Journal of Organizational Behavior, 22(2): 161-177.
43
Chandler, A. D.
1962 “Strategy and Structure: Chapters in the History of American Enterprise.” MIT Press,
Cambridge, MA.
Csaszar, F. A.
2012 “Organizational structure as a determinant of performance: Evidence from mutual funds.”
Strategic Management Journal, 33(6): 611-632.
Csaszar, F. A., and J. P. Eggers.
2013
“Organizational Decision Making: An Information Aggregation View.” Management
Science.
Cyert, R. M., and J. G. March.
1963 “A Behavioral Theory of the Firm.” Prentice-Hall: Englewood Cliffs, NJ.
Davis, G. F., and H. R. Greve.
1997 “Corporate elite networks and governance changes in the 1980s.” The American Journal
of Sociology, 103(1): 1-37.
De Figueiredo, J., and M. Kyle.
2001 “Competition, Innovation, and Product Exit.” MIT Center for Innovation in Product
Development, Working Paper.
De Figueiredo, J., and M. Kyle.
2006 “Surviving the gales of creative destruction: The determinants of product turnover.”
Strategic Management Journal, 27(3): 241-264.
De Figueiredo, J. M., and D.J. Teece.
1996
“Mitigating procurement hazards in the context of innovation.” Industrial and Corporate
Change, 5(2): 537-559.
Denis, J. L., G. Dompierre, A. Langley, and L. Rouleau.
2011 “Escalating indecision: Between reification and strategic ambiguity.” Organization
Science, 22(1): 225-244.
Denrell, J., and J. G. March.
2001 “Adaptation as information restriction: The hot stove effect.” Organization Science,
12(5): 523-538.
Edmondson, A.
44
1999 “Psychological safety and learning behavior in work teams.” Administrative Science
Quarterly, 44(2): 350-383.
Egelhoff, W. G.
1982 “Strategy and structure in multinational corporations: An information-processing
approach.” Administrative Science Quarterly: 435-458.
Eggers, J. P.
2012 “Falling Flat Failed Technologies and Investment under Uncertainty.” Administrative
Science Quarterly, 57(1): 47-80.
Eggers, J. P., and S. Kaplan.
2009 “Cognition and renewal: Comparing CEO and organizational effects on incumbent
adaptation to technical change.” Organization Science, 20(2): 461-477.
Eisenhardt, K. M.
1989 “Making fast strategic decisions in high-velocity environments.” Academy of
Management Journal, 32: 543-576.
Elsbach, K. D., P.S. Barr, and A.B. Hargadon.
2005 “Identifying situated cognition in organizations.” Organization Science, 16(4): 422-433.
Ethiraj, S., D. Levinthal.
2009 “Hoping for A to Z While Rewarding Only A: Complex Organizations and Multiple
Goals.” Organization Science, 20(1): 4-21.
Gaba, V. and J. Joseph
2012 “Corporate Structure and Performance Feedback: Aspirations and Adaptation in M-form
Firms,” Organization Science, forthcoming.
Galbraith, J. R.
1974 “Organization design: An information processing view.” Interfaces, 4(3): 28-36.
Gavetti, G.
2005 "Cognition and Hierarchy: Rethinking the Microfoundations of Capabilities'
Development." Organization Science, 16(6): 599-617.
Gavetti, G., D. Levinthal, and W. Ocasio.
2007 “Perspective—Neo-Carnegie: The Carnegie School’s Past, Present, and Reconstructing
for the Future.” Organization Science, 18(3): 523-536.
Greenstein, S.M., and J. B. Wade.
45
1998 "The product life cycle in the commercial mainframe computer market, 1968-1982." The
RAND Journal of Economics: 772-789.
Greve, H. R.
1998 “Performance aspirations, and risky organizational change.” Administrative Science
Quarterly, 43: 58-86.
Greve, H. R.
1999 “The Effect of Core Change on Performance: Inertia and Regression toward the Mean.”
Administrative Science Quarterly, 44:590.
Greve, H. R.
2003a “A behavioral theory of R&D expenditures and innovation: Evidence from shipbuilding.”
Academy of Management Journal, 46: 685-702.
Greve, H. R.
2003b “Organizational Learning From Performance Feedback: A Behavioral Perspective On
Innovation and Change.” Cambridge University Press, Cambridge, UK.
Greve, H. R.
2007 “Exploration and exploitation in product innovation.” Industrial and Corporate Change,
16(5), 945-975.
Gulati, R.
1995 “Does familiarity breed trust? The implications of repeated ties for contractual choice in
alliances.” Academy of Management Journal, 85-112.
Gulati, R., P. Lawrence, and P. Puranam.
2005 "Adaptation in Vertical Relationships: Beyond Incentive Conflict." Strategic
Management Journal 26(12): 415-440.
Gulati, R., P. Puranam, and M. Tushman.
2012 “Meta‐organization design: Rethinking design in interorganizational and community
contexts.” Strategic Management Journal, 33(6): 571-586.
Guler, I.
2007 “Throwing good money after bad? Political and institutional influences on sequential
decision making in the venture capital industry.” Administrative Science Quarterly,
52(2): 248-285.
Hannan, M. T., and J. H. Freeman.
46
1984 “Structural inertia and organizational change.” American Sociological Review, 90: 149164.
Harkness, J., B. E. Pennell, and A. Schoua‐Glusberg.
2004 “Survey questionnaire translation and assessment.” Methods for testing and evaluating
survey questionnaires, 453-473.
Henderson, A. D., and I. Stern.
2004 “Selection-based learning: The coevolution of internal and external selection in highvelocity environments.” Administrative Science Quarterly, 49:39-75.
Henderson, R., and I. Cockburn.
1994 “Measuring Competence? Exploring Firm Effects in Pharmaceutical Research.” Strategic
Management Journal,15.S1: 63-84.
Huber, G. P.
1991 “Organizational learning: The contributing processes and literatures.” Organization
Science 2(1): 88-115.
Huber, G., C. Miller, and W. Glick.
1990 “Developing More Encompassing Theories about Organizations: The CentralizationEffectiveness Relationship as an Example.” Organization Science. 1(1): 11-40.
Hu, S., D. Blettner, and R. A. Bettis.
2011 "Adaptive aspirations: performance consequences of risk preferences at extremes and
alternative reference groups." Strategic Management Journal, 32.13: 1426-1436.
Huy, Q. N.
2002 “Emotional balancing of organizational continuity and radical change: The contribution
of middle managers.” Administrative Science Quarterly, 47(1): 31-69.
Informa Telecoms and Media.
2008 “WCIS Insight: Global Mobile Market Outlook 2009.” Informa Telecoms and Media.
Ireland, R. D., M.A. Hitt, R. A. Bettis, and D. A. De Porras.
1987 “Strategy formulation processes: Differences in perceptions of strength and weaknesses
indicators and environmental uncertainty by managerial level.” Strategic Management
Journal, 8(5): 469-485.
Jaccard, J., and R. Turrisi.
2003 “Interaction Effects in Multiple Regression.” Sage Publications, London.
47
Jansen, J. J., F.A. Van Den Bosch, and H.W. Volberda.
2005 “Managing potential and realized absorptive capacity: how do organizational antecedents
matter?” Academy of Management Journal, 48(6): 999-1015.
Joseph, J., and W. Ocasio.
2012 “Multiple Goals, Selective Attention and Organizational Decision Making: A Case Study
of Technology Commercialization at Motorola,” Working Paper.
Kaplan, S. N., and B. Minton.
2006 How has CEO turnover changed? Increasingly performance sensitive boards and
increasingly uneasy CEOs (No. w12465). National Bureau of Economic Research.
Kane, A. A., L. Argote, and J.M. Levine.
2005 “Knowledge transfer between groups via personnel rotation: Effects of social identity and
knowledge quality.” Organizational Behavior and Human Decision Processes, 96(1): 5671.
Keil T, R. McGrath, T. Tukiainen
2009 “Gems from the ashes: Capability creation and transformation in internal corporate
venturing.” Organization Science 20(3): 601-620.
Kraatz, M. S.
1998 “Learning by association? Interorganizational networks and adaptation to environmental
change.” Academy of Management Journal: 621-643.
Lant, T. K.
2002 “Organizational cognition and interpretation.” Oxford: Blackwell: 344-362.
Lant, T. K., F. Milliken, B. Batra.
1992 “The role of managerial learning and interpretation in strategic persistence and
reorientation: an empirical exploration.” Strategic Management Journal, 13: 585-608.
Lave, J., and E. Wenger.
1991 “Situated learning: Legitimate peripheral participation.” Cambridge University Press.
Lawrence, P R, and J W Lorsch.
1967 “Organization and Environment.” Harvard University Press.
Levinthal, D.A, and J. G. March.
1993 “The myopia of learning.” Strategic Management Journal,14.S2: 95-112.
March, J. G.
48
1988 “Variable risk preferences and adaptive aspirations.” Journal of Economic Behavior and
Organization, 9(1): 5-24.
March, J. G., and H. A. Simon.
1958 “Organizations.” John Wiley, New York, NY.
Martin J. A., and K.M. Eisenhardt.
2010 “Rewiring: Cross-business-unit collaborations and performance in multi-business
organizations.” Academy of Management Journal, 53(2): 265–301.
Martin, J. A.
2011 “Dynamic managerial capabilities and the multibusiness team: the role of episodic teams
in executive leadership groups.” Organization Science, 22(1): 118-140.
McNamara, G., and P. Bromiley.
1997 “Decision Making in an Organizational Setting: Cognitive and Organizational Influences
on Risk Assessment in Commercial Lending.” Academy of Management Journal, 40(5):
1063-1088.
Mezias, S. J., Y.R. Chen, and P. R. Murphy.
2002 “Aspiration-level adaptation in an American financial services organization: A field
study.” Management Science, 48(10): 1285-1300.
Miller, D., and M.J. Chen
1994 “Sources and consequences of competitive inertia: A study of the US airline industry.”
Administrative Science Quarterly: 1-23.
Miller, D., and C. Dröge.
1986. Psychological and traditional determinants of structure. Administrative Science
Quarterly: 539-560.
Miller, K. D., W. R. Chen
2004 “Variable organizational risk preferences: Tests of the March-Shapira model.” Academy
of Management Journal, 47(1): 105–116.
Milliken, F. J., and T. K. Lant
1991 “The impact of an organization's recent performance history on strategic persistence and
change." In Paul Shrivastava, Anne Huff, and Jane Dutton (eds.), Advances in Strategic
Management, 7: 129-156. Greenwich, CT: JAI Press.
Nerkar, A. and S. Paruchuri.
49
2005 “Evolution of R&D capabilities: The role of knowledge networks within a firm.”
Management Science, 51(5): 771-785.
Ocasio, W.
1994 “Political dynamics and the circulation of power: CEO succession in US industrial firms,
1960-1990.” Administrative Science Quarterly. 39(2): 285-312.
Ocasio, W.
1995 “The enactment of economic adversity: a reconciliation of theories of failure induced
change and threat rigidity.” Research in Organizational Behavior, 17: 287-331.
Ocasio, W.
1997 “Towards an attention-based view of the firm.” Strategic Management Journal, 18: 187206.
Park, K. M.
2007 “Antecedents of convergence and divergence in strategic positioning: The effects of
performance and aspiration on the direction of strategic change.” Organization Science,
18(3): 386–402.
Bettis, R. A. and C. K., Prahalad
1995 “The dominant logic: Restrospective and extension”, Strategic Management Journal, 16
(1):5–14.
Rerup, C.
2009 “Attentional triangulation: Learning from unexpected rare crises.” Organization Science,
20(5): 876-893.
Ross, L., and R. E. Nisbett.
1991 “The Person and the Situation.” Philadelphia, PA.
Schoonhoven, C. B., K.M. Eisenhardt, and K. Lyman.
1990 “Speeding products to market: Waiting time to first product introduction in new firms.”
Administrative Science Quarterly, 177-207.
Schwab, A.
2007 “Incremental organizational learning from multilevel information sources: Evidence for
cross-level interactions.” Organization Science, 18(2): 233-251.
Siggelkow, N., and J.W. Rivkin.
50
2005 “Speed and search: Designing organizations for turbulence and complexity.”
Organization Science, 16(2): 101-122.
Simon, H. A.
1947 “Administrative Behavior.” New York: Macmillan.
Simon, H. A.
1962 “New developments in the theory of the firm.” The American Economic Review, 52(2):
1-15.
Sorenson, J., T. Stuart.
2000 “Aging, obsolescence and organizational innovation.” Administrative Sciemce Quarterly,
45: 81–112.
Sorenson, O.
2000 "Letting the market work for you: An evolutionary perspective on product strategy."
Strategic Management Journal, 21.5: 577-592.
Staw, B. M., L. E. Sandelands, and J.E. Dutton.
1981 “Threat rigidity effects in organizational behavior: A multilevel analysis.” Administrative
Science Quarterly: 501-524.
Stein, J. C.
1989 “Efficient capital markets, inefficient firms: A model of myopic corporate behavior.” The
Quarterly Journal of Economics, 104(4): 655-669.
Tripsas M., G. Gavetti.
2000 “Capabilities, cognition, and inertia: Evidence from digital imaging.” Strategic
Management Journal, 21:1147–1161.
Troianovski, A., and S. Grundberg.
2012 "Nokia's Bad Call on Smartphones.” Wall Street Journal. Web. 9 Dec. 2012.
Tsai, W., and S. Ghoshal.
1998 “Social capital and value creation: The role of intrafirm networks.” Academy of
Management Journal, 41(4): 464-476.
Tushman, M. L., and D.A. Nadler
1978 “Information processing as an integrating concept in organizational design.” Academy of
Management Review: 613-624.
Wiersema, M. F., and Y. Zhang.
51
2011 “CEO dismissal: the role of investment analysts.” Strategic Management Journal, 32(11):
1161-1182.
Williamson, O. E.
1985 “The economic institutions of capitalism.” Free Press, New York.
52
Table 1. Summary Statistics and Cross-Correlations (N = 1307 product-quarters)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Elevation+
Consultation+
Portfolio level - Perf. Above Aspirations+
Portfolio level - Perf. Below Aspirations+
Indicator: Phone has a direct replacement
Experience: Degree of Culling+
Device Sales (000s)+
Firm Sales (MMs) +
# of 3G phones in Portfolio
# of Smartphones in Portfolio
Average Portfolio Age+
Cumulative phone launches+
Vertical integration: Has semiconductor
# of competitors phones on market+
# of competitive launches+
+: Lagged by three quarters
Mean
3.15
3.39
0.10
(0.18)
0.08
0.96
33.39
605.70
28.38
27.89
5.95
303.64
0.06
161.34
36.96
Std. Dev.
0.98
0.94
0.45
0.25
0.27
0.47
61.05
1,850.00
11.67
10.25
0.70
61.97
0.24
22.69
9.67
1
2
3
4
5
6
7
8
9
10
11
12
1.00
-0.11 1.00
-0.11 0.02 1.00
-0.01 -0.04 0.06 1.00
-0.18 -0.04 0.29 0.05 1.00
0.40 0.11 -0.16 -0.04 -0.25 1.00
-0.01 0.06 0.04 -0.19 0.06 0.03 1.00
-0.26 -0.23 -0.01 -0.02 -0.05 -0.06 -0.12
1.00
0.59 -0.48 -0.02 -0.07 0.02 0.10 0.05 -0.36
1.00
0.62 -0.42 -0.01 -0.06 0.02 0.13 0.04 -0.40
0.97
1.00
-0.04 0.55 0.05 -0.01 0.08 -0.11 0.12 -0.25 -0.22 -0.18 1.00
0.00 0.63 0.04 0.07 0.03 0.02 0.06 -0.59 -0.33 -0.17 0.52 1.00
-0.26 -0.24 -0.01 -0.03 -0.05 -0.06 -0.12
1.00 -0.37 -0.40 -0.26 -0.59
0.11 -0.01 -0.28 -0.03 -0.51 0.45 0.00
0.10 -0.09 -0.09 -0.13 -0.06
-0.17 -0.04 -0.10 0.10 -0.21 -0.13 0.02
0.13 -0.16 -0.17 0.06 -0.01
All correlations larger in magnitude than .074 are significant at the 5% level
53
13
14
15
1.00
0.10 1.00
0.13 0.40 1.00
Table 2. Piecewise Exponential Hazard Rate Models for Product Exit
(1)
Unit Sales
1
2
3
4
5
7
6
8
9
10
11
12
13
14
15
16
17
18
19
Elevation+
Consultation+
Portfolio level - Perf. Above Aspirations+
Portfolio level - Perf. Below Aspirations+
Elevation * P>A
Elevation * P<A
Consultation * P>A
Consultation * P<A
Device Sales (000s)+
Indicator: Phone has a direct replacement
Experience: Degree of Culling+
Firm Sales (MMs) +
# of 3G phones in Portfolio
# of Smartphones in Portfolio
Average Portfolio Age+
Cumulative phone launches+
Vertical integration: Has semiconductor
# of competitors phones on market+
# of competitive launches+
Piece 3: 2Q-3Q phone life
Piece 4: 3Q-4Q phone life
Piece 5: 4Q+ phone life
N
Log-likelihood
Standardized beta coefficients
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Unit Sales Unit Sales Unit Sales Unit Sales Unit Sales Unit Sales Unit Sales
0.255*
0.222
0.484**
0.217
0.344*
-0.320**
-0.294**
-0.316** -0.486*** -0.478***
-0.718*
18.07***
2.585
19.99***
0.969***
-0.838
3.338***
2.451**
-9.193***
-9.008***
0.653***
-0.283
-0.945*
-0.727*
-0.619*** -0.572***
-0.010*** -0.010*** -0.009*** -0.009*** -0.000*** -0.009*** -0.010*** -0.009***
1.111**
1.179***
0.950**
1.012**
1.344***
1.242***
1.565***
1.551***
-0.212
-0.418
-0.0714
-0.254
-0.144
-0.358
-0.281
-0.402
-0.003*
-0.003*
-0.003*
-0.003*
-0.000*
-0.003*
-0.003*
-0.003*
-0.194
-0.333*
-0.0585
-0.181
-0.265*
-0.170
-0.266
-0.152
0.185
0.307*
0.0428
0.152
0.253*
0.150
0.237
0.133
-0.433*
-0.295
-0.491*
-0.371
-0.412*
-0.389
-0.472*
-0.523*
-0.011
-0.021*
-0.000
-0.009
-0.015
-0.007
-0.013
-0.005
14.01
14.69
17.11*
17.80*
12.59
17.65*
14.82*
15.77*
0.001
0.001
0.000
0.000
-0.000
0.001
0.000
0.001
0.016*
0.017**
0.016*
0.017**
0.009
0.009
0.009
0.009
-19.71
-13.61
-21.36
-17.95
-16.37
-18.87
-17.11
-19.53
1.369
5.081
-2.647
0.754
3.832
1.687
3.211
0.913
2.172
5.887*
-1.831
1.574
4.201
2.137
3.582
1.338
1307
1307
1307
1307
1307
1307
1307
1307
140.67
142.64
144.66
146.15
164.78
187.88
187.05
201.08
* p<0.05 ** p<0.01 *** p<0.001
+: Lagged by three quarters
54
Table 3. Models 1-7, demonstrating robustness to revenue formulation of portfolio performance-feedback variables
Piecewise Exponential Hazard Rate Models (Exit)
1 Elevation+
2 Consultation+
3 Portfolio level - Perf. Above Aspirations+
4 Portfolio level - Perf. Below Aspirations+
5 Elevation * P>A
6 Elevation * P<A
7 Consultation * P>A
8 Consultation * P<A
9 Device Sales (000s)+
10 Indicator: Phone has a direct replacement
11 Experience: Degree of Culling+
12 Firm Sales (MMs) +
13 # of 3G phones in Portfolio
14 # of Smartphones in Portfolio
15 Average Portfolio Age+
16 Cumulative phone launches+
17 Vertical integration: Has semiconductor
18 # of competitors phones on market+
19 # of competitive launches+
Piece 3: 2Q-3Q phone life
Piece 4: 3Q-4Q phone life
Piece 5: 4Q+ phone life
N
Log-likelihood
Standardized beta coefficients
(1)
Revenue
(2)
Revenue
0.255*
(3)
Revenue
-0.320**
(4)
Revenue
0.222
-0.294**
(5)
Revenue
-0.513*
0.773***
-0.010*** -0.010*** -0.009***
1.111**
1.179***
0.950**
-0.212
-0.418
-0.0714
-0.003*
-0.003*
-0.003*
-0.194
-0.333*
-0.0585
0.185
0.307*
0.0428
-0.433*
-0.295
-0.491*
-0.011
-0.021*
-0.000
14.01
14.69
17.11*
0.001
0.001
0.000
0.016*
0.017**
0.015*
-19.71
-13.61
-21.36
1.369
5.081
-2.647
2.172
5.887*
-1.831
1307
1307
1307
140.67
142.64
144.66
* p<0.05 ** p<0.01 *** p<0.001
55
(6)
Revenue
0.420**
-0.356***
9.318**
-0.601
-4.811**
0.504**
-0.009*** -0.000*** -0.009***
1.012**
1.260***
1.073**
-0.254
-0.190
-0.309
-0.003*
-0.000*
-0.003*
-0.181
-0.244
-0.217
0.152
0.233
0.189
-0.371
-0.460*
-0.446*
-0.009
-0.013
-0.007
17.80*
14.49*
16.95*
0.000
-0.001
-0.002
0.017**
0.011
0.013*
-17.95
-16.025
-17.69
0.754
3.319
2.502
1.574
3.766
2.999
1307
1307
1307
146.15
161.29
177.28
+: Lagged by three quarters
(7)
Revenue
0.211
-0.433***
2.533*
2.721***
-0.901*
-0.486***
-0.010***
1.502***
-0.273
-0.004**
-0.259
0.230
-0.512*
-0.013
18.78**
-0.002
0.011
-17.04
3.211
3.651
1307
183.25
(8)
Revenue
0.304*
-0.457***
21.63***
2.038**
-4.676**
-0.205
-0.693*
-0.446***
-0.010***
1.432***
-0.333
-0.003*
-0.214
0.189
-0.570**
-0.005
16.02*
-0.002
0.012
-17.93
2.408
2.889
1307
189.87
Table 4. Replication of Table 2, Model 8, including year and firm dummies
Piecewise Exponential Hazard Rate Models (Exit)
1 Elevation+
2 Consultation+
3 Portfolio level - Perf. Above Aspirations+
4 Portfolio level - Perf. Below Aspirations+
5 Elevation * P>A
6 Elevation * P<A
7 Consultation * P>A
8 Consultation * P<A
9 Device Sales (000s)+
10 Indicator: Phone has a direct replacement
11 Experience: Degree of Culling+
12 Firm Sales (MMs) +
13 # of 3G phones in Portfolio
14 # of Smartphones in Portfolio
15 Average Portfolio Age+
16 Cumulative phone launches+
17 Vertical integration: Has semiconductor
18 # of competitors phones on market+
19 # of competitive launches+
Firm Dummies?
Year Dummies?
N
Log-likelihood
Standardized beta coefficients
(1)
Unit Sales
0.360*
-0.388**
38.96***
0.229
-9.044***
-0.301
-0.655
-0.552***
-0.010***
1.406***
-0.0648
-0.003**
7.267
-8.483
-0.268
0.080**
-3.806
-0.003
0.008
Yes
(3)
Unit Sales
0.394*
-0.322*
40.04***
0.446
-9.329***
-0.334
-0.647
-0.522***
-0.010***
1.238**
0.229
-0.003**
8.272
-9.613
-0.171
0.010***
-5.348
0.001
0.008
Yes
Yes
Yes
1307
1307
1307
208.61
203.09
211.59
* p<0.05 ** p<0.01 *** p<0.001
+: Lagged by 3 quarters
56
(2)
Unit Sales
0.307*
-0.501***
40.45***
0.312
-9.326***
-0.305
-0.731*
-0.538***
-0.010***
1.306***
-0.276
-0.003**
-0.171
0.151
-0.445*
-0.007
14.50*
0.000
0.009
.6
0
.2
.4
Density
.8
1
Figure 1. Histogram of phone life in quarters for LG, Motorola, Nokia, Samsung, and Sony Ericsson
0
5
10
15
Device lifetime (Quarters)
0.50
0.25
0.00
% surviving
0.75
1.00
Figure 2. Kaplan-Meier survival graph – each discontinuity is one quarter
0
500
1000
Analysis Time
57
Figure 3. Summary of hypotheses
Hypothesis
Expected Model Sign Graphic Depiction Support
1 Phase-out increases with the elevation of product portfolio decisions.
+
Yes
2 Phase-out decreases as the extent of within-level consultation increases.
-
Yes
3a
The elevation of portfolio decisions amplifies the extent to which aboveaspiration level portfolio performance decreases phase-out.
-
Yes
3b
The elevation of portfolio decisions amplifies the extent to which Belowaspiration level portfolio performance decreases phase-out.
+1
Partial
4a
The consultation in portfolio decisions amplifies the extent to which aboveaspiration level portfolio performance reduces phase-out.
-
Yes
4b
The consultation in portfolio decisions attenuates the extent to which belowaspiration level portfolio performance reduces phase-out.
-1
Yes
1
Sign flipped because performance < aspirations is coded as a negative value
58
Figure 4. The effect of performance relative to aspirations and elevation on product phase-out. Solid and dotted lines are where elevation
equals 3 and 5, respectively.
Multiplier of Hazard Rate
ELEVATION
3.0
10
9
2.5
8
7
E=5
2.0
6
1.5
5
4
E= 3
2
1
0.5
1
0
-0.4
-0.2
1
1.0
3
E= 3
E=5
0.0
0.0
0.0
P-A
P-A
59
0.2
0.4
Figure 5. The effect of performance relative to aspirations and consultation on product phase-out. Solid and dotted lines are where
elevation equals 3 and 5, respectively.
CONSULTATION
Multiplier of Hazard Rate
0.25
0.25
C= 3
C= 3
0.20
0.20
0.15
0.15
0.10
0.10
0.05
0.05
0.00
0.00
C=5
-0.4
-0.2
C=5
0.0
0.0
P-A
P-A
60
0.2
0.4